Monica Dessole
Researcher in HighPerformance Scientific Computing
I am a research fellow at CERN SFTEP working in the ROOT group. My research focuses on numerical and computational linear algebra, and it includes the development of efficient algorithms for scientific computing and the implementation of scalable, reusable and maintainable scientific software, primarily parallel solvers for differential equations, and methods for analyzing large matrices and datasets.
Short Bio
I received my PhD in Computational Mathematics at the University of Padova, where I also held a one year research fellowship year at the Department of Information Engineering. Before joining CERN, I was part the HPC/Cloud group at Leonardo Labs, where I was involved in the design and implementation of cloud stacks for Big Data analytics. Previously, I have worked on the massively parallel solution on GPUs of linear systems arising from differential equations such as highfidelity numerical simulations of noncompressible unsteady NavierStokes equations and BVODE problems. In particular, I have explored the performance attained by the approximate iterative solution on GPUs of sparse triangular linear systems in the context of ILU preconditioning with application to nonmiscible dual fluid flow simulations. Moreover, I developed a direct GPU solver tailored for structured matrices arising e.g. from the discretization of two point boundary value problems. I have also dealt with the efficient solution of illposed problems my means of lowrank models and statistical techniques with application to analysis of datasets in high dimensional spaces. I developed a block pivoting technique which can be adopted to increase performance in rank deficient QR computations and in sparse recovery problems.
For further information please check out my cv.
Teaching
MOOC: Scientific Computing in Python (in Italian)Introduction to Python ( Italian version )
Software
LHDM  Activeset solver for NonNegative Least Squares based on Deviation Maximization column selectionQRDM  QR factorization for rankdeficient matrices based on Deviation Maximization column selection
dCATCH  Nearlyoptimal polynomial regression designs on highdimentional point clouds
PARASOF  Direct GPU solver for linear systems with BABD structure
ns2div  Hybrid CPUGPU solver for NavierStokes equations with variable density
Publications
Preprints

GenVectorX: A performanceportable SYCL library for Lorentz Vectors operationsArXiv Preprint, 2023[ link ]
Conference Papers

Migrating CUDA to SYCL: A HEP Case Study with ROOT RDataFrameIWOCL '24: Proceedings of the 12th International Workshop on OpenCL and SYCL, 2024[ link ]
Journal Papers

Accurate detection of hidden material changes as fictitious heat sources from thermographic dataNumerical Heat Transfer, Part B: Fundamentals, 2023[ link ]

The LawsonHanson Algorithm with Deviation Maximization: Finite Convergence and Sparse RecoveryNumerical Linear Algebra with Applications, 2023[ link ]

dCATCHA Numerical Package for dVariate Near GOptimal Tchakaloff Regression via Fast NNLSMathematics, 2020[ link ]

A massivelyparallel algorithm for Bordered Almost Block Diagonal systems on GPUsNumerical Algorithms, 2020[ link ]

Accelerating the LawsonHanson NNLS solver for largescale Tchakaloff regression designsDolomites Research Notes on Approximation, 2020[ link ]

Fully iterative ILU preconditioning of the unsteady NavierStokes equations for GPGPUComputers & Mathematics with Applications, 2019[ link ]
PhD thesis
 Topics in Numerical Linear Algebra for High Performance Computing, University of Padova, 2022